$278B of Burn vs. $122B Raised: OpenAI's 2027 IPO Just Became a Funding Test

On September 18, 2026, the Financial Times published an investor presentation from OpenAI projecting $278 billion in cumulative negative free cash flow between 2026 and 2030 — a five-year cash drain on a scale no private company in history has previously committed to in a single document. The same presentation forecasts revenue climbing tenfold from $36 billion this year to $350 billion in 2030, producing $840 billion of cumulative revenue, against $856 billion of cumulative compute and infrastructure spend.

The forecast supersedes an earlier projection The Information and Bloomberg reported in February 2026, which had OpenAI hitting $280 billion in 2030 revenue on $112 billion of cumulative cash burn. The September revision lifts revenue 25% but raises the burn envelope by roughly 2.5x, implying that operating leverage — not revenue scale — is the variable management is stress-testing.

If the FT numbers are taken at face value, OpenAI's gross unfunded need over 2026–2030 is $278 billion of free cash flow plus whatever is required to refinance the existing $122 billion cushion once it is exhausted in 2028. The cash already raised covers less than two years of forward burn. HSBC, working from OpenAI's earlier projection set, sized a $207 billion net funding shortfall by 2030 in November 2025 — a number the FT's updated revenue trajectory brings only marginally closer to balanced.

The reason the gap can close at all is that OpenAI's compute suppliers have effectively pre-committed to fund the shortfall themselves, through investments in OpenAI equity and through long-dated take-or-pay-style cloud contracts that defer cash payment into the back half of the decade. That circular structure is the non-obvious mechanism that turns a $278 billion burn into something the market can plausibly underwrite.

Three contracts anchor the loop. In September 2025, NVIDIA and OpenAI announced a strategic partnership to deploy at least 10 gigawatts of NVIDIA systems, with NVIDIA committing to invest up to $100 billion in OpenAI progressively as each gigawatt flips on. In the same period, OpenAI signed a five-year, $300 billion cloud agreement with Oracle covering 4.5 gigawatts of Stargate capacity starting in 2027. And in April 2026, Microsoft restructured its long-running partnership, dropping exclusivity on OpenAI's serving stack while keeping OpenAI's revenue-share payments to Microsoft flowing through 2030 under a hard cap.

The structure is self-reinforcing only while all three legs hold. If OpenAI misses its revenue trajectory, NVIDIA's forward order book stalls, Oracle's data-center utilization falls short of the $300B contract, and Microsoft's capped revenue share looks smaller than forecast. Each vendor is exposed to the same single-bullet revenue risk that the FT forecast is built around.

CFO Sarah Friar told employees in August 2026 that OpenAI \"will be a public company in 2027,\" and reporting in The New York Times and Forbes points to a target valuation near $1 trillion. The FT's new forecast is the first time a specific cash-burn schedule has been attached to that listing window, and it changes the question public-market investors will be asked to answer: not \"what is OpenAI worth?\" but \"how much new equity must OpenAI sell before and at IPO to reach the 2030 finish line still solvent?\"

At a $1 trillion target valuation, a single 10% primary issuance would clear $100 billion of forward funding need — enough to cover roughly one year of the FT's projected burn. To fund the full gap before 2030, the IPO and any pre-IPO rounds together would need to raise more than 20% of the listing valuation in fresh equity, a dilution profile closer to a Series H than a public-market debut.

The forecast's most fragile assumption is the tenfold revenue ramp. OpenAI disclosed $40 billion in annualized run-rate revenue by August 2026, per company statements, so reaching $36 billion in calendar 2026 is consistent with current trajectory. Hitting $350 billion in 2030, however, requires roughly 57% compound annual growth from the 2026 base — fast even by hyperscaler standards, and roughly double Microsoft's revenue CAGR over the past five fiscal years.

Operating leverage is the lever that makes the math work even when revenue comes in light. Per HSBC's earlier analysis, OpenAI's 2030 model assumed $140.7 billion of cumulative R&D and $75.4 billion of cumulative cost of goods sold against $153.8 billion of revenue — a gross-margin trajectory that the September forecast implies will be at least 25% richer. Whether the model AI business can compress inference cost fast enough to clear that bar is the open question the FT forecast does not answer.

OpenAI's projected $856 billion in cumulative compute spend is the single largest demand signal in the AI build-out. It is also the demand signal that most concentrated on three vendors. NVIDIA captures the chip layer; Oracle captures the hyperscale lease layer; Microsoft captures the application, distribution, and hybrid-cloud layer. The FT forecast implicitly makes each of them a leveraged bet on OpenAI's revenue trajectory holding.

Downstream, the same forecast shapes a thinner set of beneficiaries. OpenAI disclosed 9 million paying business users and roughly 50 million consumer subscribers by mid-2026, with enterprise representing more than 40% of revenue. The downstream pipeline that captures ChatGPT-embedded distribution — Salesforce, ServiceNow, and a long tail of enterprise software vendors monetizing OpenAI inference — is exposed to the same revenue ramp, but with the additional variable of API pricing continuing to compress.